Report on processing for sampling L8 tessels

This document is used to report results of processing focused on automatically sampling L8 tessels for fireCCI validation activities.

Input data

Data used are:

  • 2018 fireCCI total burnt area raster (0.25° spatial resolution)
  • L8 tesselation, with biome codification
  • Results of analysys of length of continuos L8 time series, with 16 days maximum “step” (Work by Matteo Sali). This dataset shows, for each L8 tessel, the maximum length of a 2018 time series of L8 images satisfying the following conditions:

    1. Maximum difference between consecutive dates below or equal to 16 days
    2. Maximum cloudiness below 30%, for all images

Classification of tessels in High an Low fire intestity

For each tessel, we first of all computed the total 2018 burnt area starting from the fireCCI raster.

we then computed the 20th percentile of the distribution of cumulated burnt area, and used it to classify tessels in High and Low intensity areas:

Analysis of cardinality of the groups

Information about Fire Intensity and length of continuous L8 serie for each tessel was joined in a single dataset to allow an analysis of cardinality of the different groups.

First of all, we computed the frequency histograms, for each biome as a whole:

, and divided as a function of intensity:

This shows clearly that in some biomes, the “High Intensity” tessels are very few, in particular for higher values of time seriers length. This can lead to problems when sampling.

To understand the relationship between cardinality of the different groups and length of the continuous series, we finally computed the cumulated number of L8 tessels showing a continuous time series above a specified threshold. Also in this case, the analysiss was done both for each biome as a whole:

, and separatedly by High / Low fire intensities:

It is evident that for most biomes, the number of High Intensity tessels with a continuous lenght above 50 days is really small.

In the map below, you can see how the number and location of available tessels vary as a function of biome and length of the considered continuous time serie.